Many thanks for checking the logs.
One thing I did notice is that the 017 log for the most recent bootloader has two extra lines regarding the SPI module:
[ 12.182517] fbtft_device: Deleting spi0.1 (spi0.1)
[ 12.219806] fbtft_device: Deleting spi0.0 (spi0.0)
which are the SPI devices for stmpe610 and fb_ili9340. Could that be important?
I've also tested the two different bootloaders on a different Pi, running a different project with another TFT of the same model, and am seeing a 10% slowdown when profiling the screen update.
Please would it be possible for you to try and see if you can replicate this...or not? I appreciate that the original TFT code comes from a third party, and that the bootloader package contains a complete kernel - but knowing if this performance and noise issue can be found by someone else independently is useful for working out how to resolve this. Note that you have to use a model B rather than a B+, as the 0421 bootloader predates the B+.